Achieving secure communications in networks has been one of the most important problems in information technology. Dolev, Dwork, Waarts, and Yung have studied secure message transmission in one-way or two-way channels. They only consider the case when all channels are two-way or all channels are one-way. Goldreich, Goldwasser, and Linial, Franklin and Yung, Franklin and Wright, and Wang and Desmedt have studied secure communication and secure computation in multi-recipient (multicast) models. In a ``multicast channel'' (such as Ethernet), one processor can send the same message--simultaneously and privately--to a fixed subset of processors. In this paper, we shall study necessary and sufficient conditions for achieving secure communications against active adversaries in mixed one-way and two-way channels. We also discuss multicast channels and neighbor network channels.